In a contemporary, diverse cohort of 9513 patients with AF considered at increased stroke risk receiving care within an integrated health care delivery system, 72% were receiving guideline-recommended use of OAC. Furthermore, manual review of medical records in a random sample of patients not receiving OAC suggests only a small proportion of such patients were eligible or were willing to be anticoagulated. This study highlights how an integrated health care system, such as Kaiser Permanente, can positively impact the treatment of chronic conditions like Atrial Fibrillation.

Rates of anticoagulation in our population are higher than reported in the National Cardiovascular Data Registry-PINNACLE Registry, 57% (NCDR-PINNACLE) [17], and similar to the smaller Outcomes Registry for Better Informed Treatment of Atrial Fibrillation, 76.2% (ORBIT-AF) [11]. The NCDR-PINNACLE noticed a plateau effect of OAC penetration among eligible AF patients considered at higher stroke risk (i.e., CHADS2 scores ≥2) [3]. Our data demonstrated numerically higher prescription rates of OAC therapy across higher CHADS2 scores, with a plateau effect evident at a CHADS2 score of 4. Given that the patients with elevated CHADS2 scores are at the highest predicted risk for thromboembolic events, appropriate treatment with OAC has a greater absolute benefit in reducing stroke burden among these patients. These findings may be attributable to the efficiencies of an integrated healthcare delivery system, which allows not only for rapid multispecialty communication, but access to care and reduced reimbursement barriers.

Dabigatran is currently the preferred OAC choice for stroke prevention in AF at KPMAS. In contrast, the OAC therapy in the NCDR PINNACLE and ORBIT-AF was predominantly warfarin, while the subsequent ORBIT-AF II registry established in 2014 reported OAC rates of 22% warfarin and 41% rivaroxaban, reflecting a shift towards the convenience and possibly lower risk profile of the newer DOAC classes. Interestingly, our data reported 33.8% warfarin use, possibly reflecting the burden of comorbidities in our population (CKD, advanced age) which would preclude the safe use of dabigatran. Despite the higher use of warfarin, appropriate anticoagulation rates remained relatively high as reported in our study.

Female sex is an independent predictor of ischemic stroke risk in patients with AF off OAC [9, 11, 18], and results of reported appropriate OAC use by sex are mixed. Previous research including a retrospective study of Medicare beneficiaries [8] and a recent study found that females have lower initiation rate of any OACs, including DOACs in newly diagnosed AF compared to males [9]. In the ORBIT-AF registry of 10,135 patients at 176 US sites, 42% were female, and females had similar OAC rates to men despite having more functional impairment and lower self-reported quality of life than men [11]. Our study results demonstrated anticoagulation rates were not significantly different by sex, possibly attributable to the effects of an integrated health care system.

Prior data on racial disparities indicate that Black and Hispanic patients [12, 19] as well as American Indian and Alaskan Native patients may have lower rates of OAC compared to White patients [14]. In addition, a large-scale analysis reported higher rates of stroke and death in Black and Hispanic patients with AF [19, 20]. We found that Black and Asian patients were significantly less likely than White patients to receive guideline-recommended OAC. This is consistent with prior studies showing a disparity in anticoagulation use in underrepresented racial/ethnic groups. There was no significant difference in anticoagulation rates between Hispanic and White patients in our study, although the number of Hispanic patients was relatively low.

When adjusted for CHADS2 scores, the observed differences in OAC use only persisted for Black and White patients with lower CHADS2 scores of 2 or 3. In contrast, continued disparity in anticoagulation rates was seen between Asian or Pacific Islander and White patients with higher CHADS2 scores of 5 or greater. These findings may reflect the role of ethnicity in affecting patient behavior, healthcare literacy, as well as the level of comorbid conditions in these groups, which could drive decisions about anticoagulation, or patient willingness to initiate OAC. Further study is needed to delineate those differences to ensure the equitable access to AF management and treatment across all ethnic groups.

Importantly, our cross-sectional study included significantly higher rates of Black patients (35.5% of total AF patients) than were included in some other registries; the ORBIT I and ORBIT II studies included only 5.0 and 4.9%, respectively [21], and NCDR Pinnacle included 2.9% [22]. This is also true for Asian or Pacific Islander patients of which there were 0.6% in NCDR PINNACLE [22] as compared with 8.3% included in our analysis. Ensuring adequate representation of racial/ethnic groups even in observational studies is critical to better understanding differences in treatment and outcomes between these groups.

In most studies investigating epidemiological disparities in anticoagulation rates for AF patients, access to care and socioeconomic factors have been proposed barriers to standard care, especially the cost of DOACs [15, 23]. In addition to lower likelihood of treatment with DOACs for AF [15], lower socioeconomic status has been associated with poorer clinical outcomes [23], lower health related quality of life [24], and lower rates of catheter ablation [25]. In contrast, our SES analysis showed no difference in rates of guideline-mandated anticoagulation among all SES quartiles. The SES-SVI groups were well balanced numerically, which suggests that while surprising, this may not be a chance finding. It may be due to the unique model of KP as a simultaneous private insurer and provider, with many patients being on Medicare and eligible for low-cost preferred formulary medications.

There has been limited contemporary data published on the reasons for lack of anticoagulation in eligible patients with AF. Prior studies have typically surveyed patients and/or prescribers for possible barriers [23, 26, 27], and such barriers include medication cost, access to care, and comorbidities conferring an increased risk of bleeding. The most frequent reason for lack of anticoagulation risk in our AF population was a history of prior or current bleeding or elevated bleeding risk. Patients being lost to follow up and patient refusal were next in frequency, and we noted a high burden of comorbid psychiatric illness. A significant number of patients had one or more barrier to anticoagulation, and importantly, only a small proportion of patients not receiving guideline-recommended OAC were eligible or willing to be anticoagulated. Larger population analyses are needed to characterize these barriers on a broader and more generalizable scale.

Our study has several limitations. We were restricted to using CHADS2 score when analyzing our study population given the current limits of the AF Registry; CHADS2VASc or ATRIA would have been better choices given more accurate stroke risk prediction with these scores. Bleeding risk among different subgroups was not assessed in this study, which may be an unmeasured confounding factor. Additionally, our patient experience may not be widely generalizable to all populations and practice settings. The KP model is a closed group-model, integrated healthcare delivery system, with patients receiving care nearly exclusively from Permanente physicians, and with a direct and efficient referral system between primary care and cardiology specialists. Access to care and medication cost is less of an issue for private insurance holders with KP given that that nearly all members have a pharmacy benefit with low drug co-pays. As this was a retrospective cross-sectional study, some recently diagnosed patients with AF may have been counted as not appropriately anticoagulated even though they may not have yet had the opportunity to complete their evaluation for OAC eligibility. Finally, our anticoagulation regimens with dabigatran as the preferred DOAC may not be broadly generalizable to all other U.S. practice settings, where the most common DOAC currently prescribed is apixaban [28].